looks like it is failing because numpy got upgraded to version 2.3.4 on October 15

looks like it is failing because numpy got upgraded to version 2.3.4 on October 15

Trying to fix in #1433

pyarrow 21.0.0 cannot be built with python3.14. We'll need to wait for apache/arrow#47438.

Correct @cmp0xff pyarrow is not yet compatible with 314 so will have to wait for that, I tried a few days ago and fell into the same trap.

I think the failures are now all FutureWarnings. How should we proceed?

Need to understand why the warnings happen with 3.14, but not with earlier versions of python.

Can you create a simple example that doesn't show the warning with 3.13 but does show it with 3.14 and then create a pandas issue?

It only raises on 3.14, I dug into it and turns out it will alert if sys.getrefcount(self) in https://github.com/pandas-dev/pandas/blob/f4e1c73159cc59a5aff076604f8b43a08c4dbf5d/pandas/core/frame.py#L4291 is under 3, it is exactly 3 in 3.14 but 4 in 3.13.
Maybe a bit painful to wrap all the tests, but the warning is at least raised inside the functions and not in our tests so open to options

It only raises on 3.14, I dug into it and turns out it will alert if sys.getrefcount(self) in https://github.com/pandas-dev/pandas/blob/f4e1c73159cc59a5aff076604f8b43a08c4dbf5d/pandas/core/frame.py#L4291 is under 3, it is exactly 3 in 3.14 but 4 in 3.13. Maybe a bit painful to wrap all the tests, but the warning is at least raised inside the functions and not in our tests so open to options

Can you report this on the pandas repo?

I did a little digging and turns out the pandas team is aware: pandas-dev/pandas#61368

I did a little digging and turns out the pandas team is aware: pandas-dev/pandas#61368

Great!! Let’s wait until the issue is resolved there.